“DON'T KNOW” RESPONSES TO RISK PERCEPTION MEASURES: IMPLICATIONS FOR UNDERSERVED POPULATIONS

Purpose: Risk perceptions vary by multiple sociocultural factors that are associated with both numeracy and health disparities (e.g., race/ethnicity).  Variations in the wording and format of risk questions can also affect responses.  However, few risk perception measures include a “don’t know” option.  Assessing how “don’t know” responses relate to sociocultural factors may help identify potential challenges in measuring perceived risk and help elucidate how diverse populations conceptualize health risk information.   

Methods: Data from the 2005 National Health Interview Survey were analyzed (N=31,202); only participants age 18+ with no prior cancer diagnosis were included.  Perceived comparative risk of colorectal cancer was assessed.  Although not an explicit response choice, 8.1% (N=2,190) of the 29,122 participants who answered the perceived risk question indicated “don’t know.”  Chi-squared and t-tests were used to examine the relationship between answering “don’t know” and several sociocultural variables associated with cancer disparities.  Multivariate logistic regression examined the occurrence of “don’t know” responses based on those variables that were statistically significant in the univariate analyses.   

Results: Univariate analyses identified multiple characteristics that were significantly associated with responding “don’t know.”  Most (7 of 11) remained significant in multivariate analyses.  Being older (vs. younger), Black/African American (vs. White), never married (vs. married or vs. divorced), not born in the U.S. (vs. born in the U.S.), having less than a 4-year degree (vs. having a 4-year degree), and not reporting a family history of colorectal cancer (vs. reporting a family history) were associated with increased odds of responding “don’t know.”  Analyses were repeated using the 2005 Health Information National Trends Survey and a dataset comprised of inner city racial/ethnic minorities.  Although statistical significance was not consistent across all variables and all datasets, the overall pattern of responses were similar.   

Conclusions: Participants belonging to several historically disadvantaged groups were more likely to indicate that they did not know their risk of colorectal cancer.  This suggests that measures of perceived risk that do not include a “don’t know” response option may not adequately capture risk beliefs among some members of underserved populations.  This is especially problematic for interventions that target perceptions of risk.  Research investigating the cognitive and behavioral implications of being unwilling or unable to provide a risk estimate on receptivity to health information and behavior are needed.